AARNE NYYSSÖNEN is Professor of Forest Management and Surveys, University of Helsinki, Finland.
One chapter of a report by an FAO André Mayer Fellow
IN EARLIER times, forest inventories were carried out entirely on the ground. However, the prospects of aerial photography have aroused interest for some 40 years. Since the second world war in particular, aerial photographs have been employed on an increasing scale and in many contexts are already established as a method. There are areas, especially the northern European countries (Finland, Norway and Sweden) where the primary value of aerial photographs, because of special conditions, is still only in the form of maps facilitating the work. On the other hand, there are regions such as North America where a considerable saving has been achieved by the use of aerial photographs, in particular as a result of reduced ground work which involves high expenses. In surveying extensive areas, photographs have offered substantial advantages for the same reason.
It is particularly important to reduce the ground work required in tropical forests, which is usually both laborious and expensive because of the character of the forests, the climate, and the lack of communications. For this reason, greater hopes are attached to aerial photographs in the tropics than in the temperate zones. This makes it natural to consider here the possibilities of aerial photography in taking an inventory of the growing stock. It necessitates a detailed analysis of the questions that can be answered from photographs.
The information that can be obtained from aerial photographs is either directly visible from them or can be assessed indirectly. The present work concentrates almost exclusively on the former. Among the latter is, for example, the important matter of site quality. Being the sum of the factors of the environment that influence tree growth, it is measurable on aerial photographs to the extent that the key factors of the environment can themselves be recognized. Tree growth is a function of local climate and soil. Local climate and soil moisture are apt to be closely related to the topography, and topographical data can be classified accurately from the stereoscopic image. Site classification based on topographical differences has proved valuable in some regions (Harrison and Spurr 1955). Much progress has also been made in the identification of soils from aerial photographs. For tropical forests, however, site classification from photographs is largely an unexplored matter. Additional research is required before it can be decided what other characteristics can be interpreted from aerial photographs.
The following points will be discussed in detail below: tree species identification, classification of forest cover types, and quantitative estimation of the growing stock. From these points and a consideration of the quality of photographs, it is possible to assess the value of aerial photographs as a source of information in forest survey.
A characteristic of tropical forests is numerous tree species of varying value. One of the most important aims of forest inventories is to show the economically important incidence of valuable tree species or categories of tree species. Hence the value of aerial photographs depends largely on the extent to which they can be used to identify tree species. This can perhaps be assessed from a review of investigations in the 1950's into the identification of tree species in the tropics.
Paijmans (1951) worked from excellent photographs, scale 1:10,000, in the Malili forests of the Celebes (Sulawesi). He found that tree species could be identified in photographs in exceptional cases only. The exceptions were Anthocephalus macrophyllus in lowland forest and the Campnosperma sp. in the hills. Apart from these, the aerial photographic appearance of the crowns varied so much with differences in age and, further, usually differed so slightly from the appearance in the aerial photograph of the habitue of crowns belonging to other species that it was not reliable enough for species identification. Since no more than these two species could be identified of the approximately 100 in question (Boon 1956), the result was not encouraging. Paijmans believed, however, that photographic methods especially adapted to the needs of forestry might give better results.
Hannibal (1962) discussed the identification of tree species in Indonesian conditions on a more general level. Identification of individual tree species in the dryland forests can be seen from the foregoing. In addition, from the investigation carried out in the Malili forests, he concluded that even if any single species customarily formed groups and, when this did happen, an important part of the total crown area of all the dominant trees was represented by individuals of that species, its identification on the aerial photographs was still possible only in exceptional cases. If, however, a tree species occurred in pure stands it might be possible to identify it. This was proved in a forest in Sumatra where pure stands of Camphor (Dryabalanops sp.) could be identified although the scale of the photographs was 1:40,000.
In mangrove and in marshland forests, the situation seems to have been relatively slightly better. In mangrove forests pure stands of bakau (Rhizophora), nipah (Nipa fructicans), nibung (Oncosperma filamentosa), etc., have been identified. Likewise pure stands of gelam (Melaleuca) and sage (Metroxylon sp.) have been identified in marshland forests. It must be emphasized that these species were in pure stands. The individual tree species could not be identified since no individuals were recognizable by their size and/or appearance.
Experiments for interpretation of mangrove species in North Borneo have been done by E. C. Francis (1955). Using ground photographs in connection with 1:25,000 aerial photographs, he described the appearance of approximately 10 mangrove species from aerial photographs.
Boon (1955-56) referred to experiments in New Guinea. Although some tree species with very individual crown shapes, such as coniferous species like Agathis and Araucaria, could be identified in the photographs; most tree species could not.
In the Thailand evergreen forests, yang (Dipterocarpus alatus) is the only species which is distinctly recognizable from photographs owing to its large crowns and brightly shining leaves (Loetsch 1957a). Fortunately, yang is also the most valuable species in these forests. In deciduous forests teak can be recognized provided the photographs are taken between 15 July and 15 August (flowering time).
In southeast Asia, in the dry dipterocarp forests of Cambodia, Dipterocarpus intricatus can be identified on photographs thanks to its light color.
Two tree species could be identified on the aerial photographs in the Sinharaja forest in Ceylon (Merritt and Ranatunga 1959). Hora (Dipterocarpus zeylanicus and D. hispidus) could be recognized when they occurred in groups, and thiniya (Doona congestiflora) both in stands and as an individual.
From his experience in Africa, D. A. Francis (1967) stated that tree species identification was only possible in a relatively limited number of eases. On panchromatic photographs of a forest area in Ghana the only recognizable species were silk cotton (Ceiba pentandra) and the umbrella tree (Musanga smithii), neither of which are of any commercial value or indicators of a valuable forest type, though musanga is usually a sign of regenerating forest.
Uganda experiments have been summarized by Cahusac (1957). Using the standard small-scale cover it was possible to identify definitely about six individual species by their crown structure and tone. In two cases the identification was of single individuals, but the other four concerned the gregarious aspect. Once identified, however, it is often possible to locate individual specimens elsewhere. Unfortunately not all the species identifiable are of value as key species for type determination.
Reporting his experience in South America, Heinsdijk wrote in 1952 that the real difficulty is in interpreting single trees. After more than two years of intensive photographic study - about half of this time was spent outdoors - Heinsdijk decided that in certain circumstances it was possible to identify a number of tree species of the Surinam forests. Later (1957-58), he stated that the identification from aerial photographs of tree species from tropical forests was very difficult and highly speculative, especially when using the scale 1:40,000. Identification with certainty was only possible in a few cases: when the upper story of the forest was dominated by one species, so that it appeared on the photograph as a pure stand (Mora excelsa, for example), or a mixed tropical forest with patches of upper story trees composed of a single species (Goupia glabra, Hymenolobium petraeum, etc.). Accurate identification was also possible when a certain forest type had only two or three species with distinctly shaped crowns in the upper story, as was usually the case with swamp forests (cf. below). These cases were exceptional in dry land forests. The author described a method, however, which made it possible to decide the composition of the forest from the biggest crowns.
Aerial photographs have been very helpful when surveying baboen (Virola surinamensis) in Surinam (Jaarverslag 1957). Tree species had to be identified from aerial photographs as far as possible, particularly those that grew on swamp land difficult to survey from the ground. Baboen grows in the main canopy; its only companion in this canopy is matakki (Symphonia globulifera). For study purposes, test photographs were taken from the air by hand, and subsequently it proved wiser to use aerial photographs rather than aerial surveys without photography. A total of some 41,000 trees could be mapped in detail by marking them with dots.
Swellengrebel (1959) recently published a paper based on an investigation in British Guiana using 1: 10,000 aerial photographs. The species of individual trees could not be identified. This was mainly because of the great variation of appearances assumed by the crowns of one tree species on the photographs. Some trees had old leaves and looked dark, others of the same species were just growing new leaves which produced light tones on the photographs. Another difficulty was that trees protruding from the general canopy usually showed up lighter than other trees of the same species. However, some species occurring in groups could be recognized on the photographs. These were more (Mora excelsa), morabukea (Mora gonggrijpii) and greenheart (Ocotea rodiaei). But even then the group was recognized by its general appearance rather than by the features of individual trees within the group. In other words, although it was known that the majority of the trees in such a group consisted of a certain species, it was very difficult to name any one tree within the group with certainty. Wallaba forest could be recognized as a type, but here again the single wallaba tree could not be recognized with certainty.
To summarize the papers quoted, identification of tree species from aerial photographs in the tropics has led to important results in some cases. Among the successfully identified tree species are, for example Dipterocarpus alatus in Thailand, Virola surinamensis in Surinam, some mangroves, etc. In those cases aerial photographs decisively facilitated the inventory.
On the other hand, the authors cited, and the opinion formed by the present author during his trips concurs, show that the results are not satisfactory in the majority of tropical forests, and that successful cases are rather exceptional. Some tree species could best be recognized when they occurred in groups but this happens relatively seldom in the tropics.
This general conclusion seems rather negative, but it agrees, for example, with the experience of species identification in eastern mixed forests of the United States of America, forests which are on the whole easier to survey than the tropical. Identification was not possible at scales of 1:15,000 to 1:20,000. But at 1:1,200 the identification was 37 percent correct, and at 1:4,800, 23 percent correct, when panchromatic film was used (Rogers 1958). The latter two scales, however, cannot be considered economical for general use.
Is progress in the near future likely to bring about an essential change in the situation?
It cannot be denied that a great deal of research remains to be done. Perhaps the most important subject is the utilization of tones, making spectrophotometric measurements like Backström and Welander (1948, 1953), Hindley and Smith (1957), and Belov and Arcybasev (1957), who achieved significant results. Success is greatest, however, when few tree species only are to be identified. In the tropics tree species are usually numerous, and they are mostly broadleaved trees. They are harder to distinguish from one another than from coniferous trees. The intraspecies variation, also emphasized in the papers by Hannibal (1952) and Swellengrebel (1959), must also be borne in mind. Hindley and Smith (1957), in particular, concluded that, owing to wide variations in spectrophotometric values within species and comparatively small variations between them, methods based on tonal differences of foliage through various film/filter combinations are likely to be of small importance in species identification from aerial photographs. The difficulties involved in species identification are not reduced by the fact that in the dense tropical forests not nearly all the trees are visible in the photographs.
There is no doubt that to gain the optimal use of aerial photographs in tropical forestry the forester should be able to identify individual tree species. The review given here is not exhaustive on the tree species problem. However, it is obvious that tree species in the tropical forests are usually not discernible from aerial photographs taken at usual scale and hence at reasonable economic cost. This is one reason why ground surveys are indispensable. It will be seen later in this paper, however, that aerial photographs have certain other important functions in tropical forest surveys.
Tropical forests consist of almost endlessly variable communities of plants. If they are to be viewed for forestry purposes, the forests must be classified in a certain way. Assessment of the surface areas of the classes formed, which in forest inventories are often termed strata, can be important in itself.
The intention now is to study in greater detail the uses of aerial photographs for classification of forest cover types. In addition to species composition, the classification may take into account, for example, the density and height of the stand, crown size and appearance, local topography, moisture and soils, as well as human influence (exploitation, shifting cultivation, etc.).
It might be useful to take illustrative examples of experience in various regions with the classification of forest cover types. It is difficult to obtain a conclusive picture in this way because the illustrative material on which opinions have been advanced varies markedly. This part of the work is limited to forest types recognizable by photo-interpretation alone, but it is possible that some of the classifications referred to necessitated a fair amount of control work on the ground.
One of the early papers reporting on the identification of distinctly differing types of vegetation from the aerial photographs and, for example, the identification of the soil from vegetation, was written by Colwell (1946). He gave an idea of how mangroves, nipah palms, etc., appear on aerial photographs.
In his paper on Indonesian conditions, Hannibal (1952) distinguished two main groups, vegetation influenced by man and natural vegetation. In the former, he distinguished the following types: old secondary forest, young secondary forest, plantations, and logged forest, in addition to which there are wet rice fields and the following types: dry agricultural fields, grass and/or lalang areas; these two are often indistinguishable from each other in aerial photographs. Three types of natural vegetation (primeval forest) are normally recognizable: mangrove forests, marshland forests, and dryland forests. In the last-mentioned, subtypes can be distinguished, for example, on the basis of density, but their limits are often difficult to define because subtypes merge with each other.
The identification of forest types in North Borneo has been described in several papers. According to Howroyd (1954), it is generally simple to separate commercial dryland forest from mangrove, nipah palm, second growth forest and other variants; the exception to this is areas which have been only partially logged. The classification scheme of E. C. Francis and Wood (1954) provides for 16 classes of vegetation cover, of which 12 are predominantly woody and natural in origin. The forest is classified, according to its topography, and the more important classes further subdivided by means of the average size of the visible tree crowns. The 16 types are arranged under six main headings, broadly classified as forest of commercial value, other noncommercial vegetation and vegetation resulting from man's activities, as follows: salt water swamp forest; transitional forest; inland forest, drained; inland forest, liable to flood; cultivation; and cleared land. The scale of photography was 1:25,000 to 1:30,000. In the authors' opinion there is little doubt that more than 16 vegetation types could be elaborated from better quality photographs.
In Sarawak, most of the following main types of natural vegetation can be recognized fairly easily on small-scale aerial photographs (Browne 1968): mangrove forests, with the usual subtypes; beach forests; mixed swamp forests; elan (Shorea albida) swamp forest; padang paya, Kerangas or heath forests; lowland dipterocarp forest; riparian forest; hill dipterocarp forest; and mossy forest.
The types distinguished by aerial mapping in Papua and New (Guinea have been described by Taylor and Stewart (1958). They give a fairly detailed description of what the types looked like on aerial photographs. The following major patterns were recognized by stereoscopic study of photographs of good quality, scale 1:40,000: mature rain forest; secondary forest - regrowth pattern; grassland - regrowth - secondary forest, either on alluvium or on hilly country; fluctuating swamp pattern; mangrove pattern with four subunits; Casuarina sp. pattern; Octomeles sumatrana pattern; low level flood pattern; blast area pattern.
In the forest inventory of the northern provinces of Thailand the strata below 1,000 meters recognized by Loetsch (1957b) from photographs of an average scale of 1:48,000 were: mixed deciduous forest; semi-evergreen forest; dry dipterocarp forest; permanent nonforested area. In Loetsch's opinion, if photographs were taken in the flowering season of teak it would be possible to distinguish between the substrata teak-bearing and non-teak-bearing mixed deciduous. The evergreen virgin forest in one province could be subdivided into four strata on 1:15,000 photographs by using the incidence of yang trees (Dipterocarpus alatus) as the primary basis of division (Loetsch 1957a).
Two large-scale inventories have been undertaken in Cambodia. In the survey east of Mekong the classification applied by Mr. B. Rollet consists mainly of the following types; dense evergreen forest; mixed moist deciduous; dry dipterocarp; second growth, or plenty of shifting cultivation; savanna; inundated grassland; swamps; scrub; ricefields; rubber plantations; forest plantation; and bamboos. The first three are the main types, and there are of course also some mixed types.
Wheeler's (1959) survey plan has five forest classifications: dense hardwood; open hardwood; dense pine; open pine; and inundated mangrove. It is possible that palm, bamboo, an unclassified and a nonproductive class will be added. Nonforest classification are also needed. It is not clear from the plan whether the intention is to identify all these classes from aerial photographs or whether ground work also will be required.
Merritt and Ranatunga (1959) and de Rosayro (1959) have reported on experiments in (Ceylon. Ecological types were mapped in the 10,000-hectare Sinharaja forest, located in the wet zone, from aerial photographs scale 1:15, 840. The following number of strata were recognized mainly by photo interpretation: eight in the virgin forest, two in the secondary forest, and four nonforest classes. When the country's forests were surveyed under the Colombo Plan, a much simpler stratification was adopted in the interpretation from aerial photographs of the dry zone. On 1:40,000 photographs the forests were divided into three strata: medium productive, low productive, and nonproductive class. No attempt was made at species identification, and the classification was based mainly on crown coverage, crown diameter, and height relations.
Little has been published on the definition of forest types from aerial photographs outside southeast Asia. D. A. Francis (1957) and Cahusac (1967) reported on their experience in Africa, but without giving any classifications, and dealt with the subject in general terms. In Surinam, as a result of preliminary work by Heinsdijk (1952), six nonforest types have been recognized from aerial photographs. Further, the following forest classes were seen: mangrove forests; swamp forest with five distinctions; marshland forest; forest on dry land. In the important group of dry land forests it was only possible to see a few outstanding and easily recognizable types.
In a later paper, with a methodical analysis of the types, Heinsdijk (1955) described the technical procedure of type mapping and emphasized the importance of aerial photographs. Worth mentioning in this connection is the key for the interpretation of vegetation types on aerial photographs in northern Surinam, scale 1:40,000 (van Dillewijn 1957). It consists of paired aerial photographs for 30 typical forest types, with terrestrial stereoscopic photographs of each, and a brief description of the type and its distribution.
Forest type mapping with aerial photographs has been done to a large extent also in the Amazon valley in Brazil, but for type identification ground work was important (for example, Heinsdijk 1957). In Guatemala, aerial surveys have been made of forests where mahogany is the chief commercial species (cf. Harrison and Spurr 1955). Small-scale (1:40,000) photography limited the degree of classification, but it did permit generalized classification of the forest into seven types or associations.
Summarizing this review of the way in which it is possible to distinguish between different forest cover types from aerial photographs, the first point that attracts the attention is the large number of type classifications. This naturally follows from the different conditions in the extensive zone of tropical forests which girdle the world. On the other hand, it seems that the principal types have not always been seen distinctly enough, for which reason the classification is not the best possible for forestry purposes. The definition of terms reached within FAO in recent years offers a good starting point (cf. Haig 1958).
It is usually possible to distinguish between forest and nonforest areas in aerial photographs. They can often be divided into certain groups. Different mangrove, marshland and swamp forests can be classified into special groups, and their subtypes can be distinguished. This alone is often a decisive advantage in the inventory of forests, in the first place because it permits concentration of the ground work.
Usually, however, various dryland types are the most important. Such obvious categories as savanna and dry dipterocarp forests, hill and lowland forests can be distinguished. But especially in the areas of dense evergreen forests the progress made in classification has not been as great as could have been hoped. In these areas are often seen extensive territories which look practically homogeneous or which merge into each other without distinct borderlines, and their classification is often difficult (cf. Hannibal 1952, D. A. Francis 1957). It is obvious, therefore, that some of the work of identifying forest types must be done on the ground. This is indicated by experience, for example, in Thailand and Ceylon.
These points by no means reduce the value of aerial photographs in stratification. Types and categories important in inventories can be distinguished with their aid, although the picture obtained must be filled in and checked. The importance of aerial photographs is emphasized, for example, by Loetsch (1957a) who says that in a certain case the number of ground plots would have been at least four times as great in the forest had no stratification been made from photographs. In many cases the potential classification can be improved by using different methods. The most important of these takes into account topography by using stereoscopic interpretation as several forest types are closely related to the topography of the country (cf. Heinsdijk 1957). The importance of good stereograms deserves emphasis. Finally, a ground knowledge of the district to be classified is very important (cf. for example, Miller 1957; de Rosayro 1959).
FIGURE 4. - Transferring forest type boundaries from a single aerial photograph to base map by means of an Aero-Sketch-master mounted on a gliding support. This work is being done by the Sudanese forest officer responsible for the field work of forest inventory in the area covered by the photograph.
Before deciding on the most suitable forest inventory method, it is necessary to form an idea of how far aerial survey can be used for a quantitative description of the growing stock. This remains important in spite of the fact that supplementary ground survey is necessary because tree species cannot generally be distinguished sufficiently clearly from aerial photographs.
The principal characteristics which it might be possible to estimate from aerial photographs are tree height, crown diameter, crown coverage, and tree counts. As these same characteristics are employed as component factors in the classification of forest cover types, a certain amount of the work involved in quantitative estimation of the growing stock has already been done. The intention now is to obtain an idea of the significance of these characteristics for the estimation of growing stock volume. The volume of trees and stands can be estimated on photographs to the extent that it is correlated with the characteristics mentioned above, and in so far as the numerical values of those characteristics can be measured from photographs.
The volume of individual trees is estimated from tree aerial volume tables which give the average tree volume to be expected for a given tree height and crown diameter (cf. for example, Ilvessalo 1950). For stand volume, the stand aerial volume table shows the average stand volume when stand height, crown coverage, and sometimes also crown diameter are known.
Aerial volume tables have been worked out for several stand types of the temperate zones. Unfortunately, they have not given very accurate results in practice. For instance, the standard error of estimate in surveys based on volume tables worked out by the present author for pure stands of Scotch pine in Finland was ± 28 percent (Nyyssönen 1955). The results obtained by Nakayama (1958) in Japan were of identical accuracy. This means that volume estimates made solely from aerial photographs are only very approximate, and only sufficient for preliminary estimation of order of magnitude of timber volumes. In general, even when only an over-all estimate of satisfactory accuracy is desired, a certain amount of ground work is necessary to correct systematic errors.
The above applies primarily to conditions in the temperate zones and to tropical forests composed of one or a few tree species which can be compared, in this respect, with the forests of the temperate zones. As the average conditions in most of the tropical forests are much more complex, good results cannot be expected from aerial surveys. But it is not impossible that volumes estimated from aerial photographs there in the future can be used as a basis for the preliminary classification of forests; each of the classes so established could then be treated as a separate stratum for sampling purposes.
To plan the broad outlines of future work on quantitative estimation, the problem must be considered in some detail. The first point is estimation from aerial photographs of the characteristics affecting the volume.
Height (tree height, average stand height, etc.) is a characteristic very much used in temperate zone forests. In most tropical forests the height is not as serviceable, especially in aerial survey. The estimation of height is difficult or impossible. The height of a tree can generally only be measured when the tree top and the ground on which the tree grows are both visible on the photographs. This is very seldom the case in tropical forests where the ground is usually screened from view by the canopy or undergrowth (cf. Howroyd 1954, Loetsch 1957a, Swellengrebel 1959).
Apart from the difficulty of measuring, it should be noted that the length of clear bole is usually of greater interest in the tropics than the total height of the tree. Even in ground surveys the total height is by no means always measured. One explanation for this practice is that, for example, in dipterocarp forest the height of top-canopy trees is fairly constant (Howroyd 1954). Again, the correlation between top-height and diameter breast height (DBH) is considerably lower in tropical broadleaved trees than, for instance, in coniferous trees. Consequently, the use of the total height for mass evaluation is not very dependable (Loetsch 1957a).
In conclusion, the use of tree height as a characteristic in aerial surveys usually does not enter into question.
Crown coverage, the characteristic generally used in aerial stand volume tables in temperate zones, is again not so serviceable in tropical forests. The use of crown coverage is not unknown in the tropics (for example, Paelinck 1958), but as the measurement of crown coverage presupposes an incomplete cover, and this is seldom found in the dense tropical forests, the characteristic has its great limitations. However, it is possible that the application of the crown coverage of the largest trees by means of stereoscopic interpretation will prove useful.
Tree counts are of value primarily in open forests where the individual tree crowns are quite distinct (Harrison and Spurr 1965). Hence the number of trees as an indicator of forest density seems to be of little value in the tropics, although in some cases, for example, the number of trees visible on aerial photographs may be used (Swellengrebel 1959).
Crown diameter is the last of these characteristics under consideration for aerial surveys. It seems more suitable than the others in tropical forests. Not all crowns, of course, are visible and measurable on aerial photographs, but those of the biggest and at the same time usually most significant trees are. This is indicated, for example, by the percentages of visible trees in certain vegetation types of British Guiana computed by Swellengrebel (1959). The scale of the photographs was 1:10,000. It appeared that all the trees in the following DBH classes were visible on photos: in Wallaba forest 50 centimeters and over, in mixed forest and Mora forest 70 centimeters and over. Importance should be attached also to Paijman's (1951) statement that the separate crowns visible on the photographs belong, for the greater part to trees with timber of marketable dimensions; trees whose crowns are not visible do not play an important contribution to the total timber volume.
There are certain difficulties involved in measuring crown diameter. The edges of the crown sometimes may not be clear, the crowns of a group of trees may appear as one big crown, and conversely some big trees have an irregular crown made up of two or three compact tufts, etc. The skill of the interpreter may overcome these obstacles, at least to some extent, but more research will be needed before a general conclusion can be drawn on the adequacy of this procedure.
Measurement of crown diameter, however, is only one aspect of the problem. Another is the correlation between crown and stem, on which several studies have been made.
Paijmans (1951) found in a virgin forest in the Celebes that there was a rather strong linear correlation between crown and stem diameters.
Miller (1957) mentioned the preliminary work done by Howroyd (1954) in north Borneo and by Farrer in Tanganyika. In each case the relationship between crown diameter and DBH of a group of species was determined from measurements made on the ground. Both workers found that the relationship was linear, and so similar in trend within the group that a single line could represent all the species. The correlation coefficients in both cases were highly significant.
Heinsdijk (1957-58) studied the relationship between crown diameter and bole diameter from an extensive material. Both in Surinam and in the Amazon valley there was a close relationship between the crown diameter of the upper-story trees and the diameter of their boles at breast height. The differences between those geographical areas were relatively small.
In contrast to the above papers showing a fairly firm correlation, a paper by Hollerwöger (1954) concerning teak forests in central Java was less encouraging. A slight correlation (r = 0.33) only was found there between the crown diameter and the DBH.
All told, these reports show that a correlation usually also exists between crown and stem dimensions in the tropical forests. If it is possible to measure crown diameters on aerial photographs, the correlation between crown diameter and stem diameter is natural in itself; it goes without saying that the bigger the crown, usually the larger the stem.
If the volume of upper-story trees can be ascertained, the approximate total volume of growing stock usually follows, since there is a certain correlation between the two (cf. Heinsdijk 1957-58). In many cases, however, the structure of the growing stock is of special interest, for example, the volume of commercial tree species. As trees species usually cannot be identified satisfactorily from aerial photographs, it is important to know whether the volume of commercial species depends on the total volume or on the volume of the upperstory. If there is such a dependence, these volumes could be used in estimating commercial timber volumes.
Sampling data collected from forests of the Amazon valley by Heinsdijk (1957-58) show that there is evidently a certain relation between the total volume of the forest and its composition. Practically everywhere groups of tree species are found which grow taller in the heavier forests than in the low forests.
To throw light on this interesting question, the present author analyzed material measured in quite a different part of the world, in south Sumatra in 1959, kindly placed at his disposal for the purpose. As the same material is employed later in the present study also, it should perhaps be described first.
For a study of the sampling methods that might be used in rain forests, data of experimental measurements were collected from a fairly homogeneous stratum of dense lowland forest by the Indonesian Bureau of Air Photo Interpretation. The measurements were done in Waikambas forest, in the Sukadana area: the present writer had the opportunity of seeing this forest in September 1959. One part of the experiment consisted of measuring 120 sample plots, size 0.18, 0.20 or 0.22 hectare; their location in the area is shown in Figure 1. The width of the recording units was always 20 meters, and the length 90, 100 or 110 meters, respectively.
All the trees with a diameter at breast height of 25 centimeters or more were numbered and both the diameter and the length of clear bole were measured. The stem volume was then calculated, using the average form factor 0.7. The trees were divided into two groups: commercial and noncommercial. A classification like this is naturally always a matter of opinion, but the aim was to separate the tree species usually handled in commerce today from the others. The total number of tree species was about 125, about 20 of them commercial.
For analysis of the material, the volume of all the trees with a diameter of more than 42.5 centimeters was computed by sample plots; this is the usual limit of measurement in the area. The volume of the commercial trees of the same size was calculated. The volumes, each per hectare, are plotted in Figure 2. There is a very positive correlation between the two volumes. The correlation coefficient is r = 0.865, which is highly significant for so many plots. The per hectare mean total volume is 105, the mean commercial volume 67 cubic meters.
In other words, if the volume of all the trees above a certain size limit can be studied from aerial photographs, some indication can be obtained at the same time of the structure of the growing stock and at least of the amount of commercial wood.
These observations suggest the points deserving attention in future studies of quantitative estimation of growing stock from aerial photographs. It seems, however, that even in the best case the results can only be used to divide the growing stock into more homogeneous strata. Obviously, growing stock estimation from aerial photographs is still a very uncertain procedure in most of the tropical forests. The difficulty of identifying the commercial tree species, defects in the stems, etc., limits the application of direct estimation of growing stock volume from aerial photographs.
FIGURE 5. - An essential part of forest inventory work in tropical countries is the compilation of volume tables for tree species. Here a Podocarpus tree has been felled by an inventory team and is being measured for this purpose
The most important factor influencing the value of interpretation of aerial photographs is the skill of the interpreter. But the results achieved also depend greatly on the equipment, materials and methods used in the various phases of photographing, processing, and interpretation work. The season of photography, time of day and weather conditions also affect the results. All these points must be considered. It is not possible to go into details in this connection, nor is it necessary as there are various textbooks and manuals available (for example, the manual of photographic interpretation, 1960). But there might be some point in trying to draw conclusions on a couple of important aspects which have given rise to differences of opinion. They are the scale of the photographs and the quality of the film used in aerial photography.
Large-scale photographs generally facilitate detailed interpretation. But the cost is a factor here. Costs rise steeply when the scale is enlarged. The relative costs of two sets of prints and the photo index, including the cost of photography, for a large area, were reported by Harrison and Spurr (1955):
It might be worthwhile noting here the more than threefold difference between the extreme values, which shows that the smallest possible scale should be selected - it also increases the pace of the work.
The literature contains various views on scales of photography in tropical forests. Hannibal (1952) recommended a scale of 1:10,000 to 1:13,000, or at the most 1:20,000, for good photography in Indonesian conditions. Similar relatively large-scale photography has often been employed and recommended elsewhere. Swellengrebel (1959) based his study on 1:10,000 photographs. For his thorough study of the ecological significance of stratification in Ceylon, de Rosayro (1958, 1959) favored 1:10,000 to 1:15,000. In Surinam, only 1:10,000 was adequate for the photographic estimation of Virola surinamensis (Jaarverslag 1957). Although Loetsch (1957a/1957b) used 1:40,000 and even 1:48,000 photography in Thailand, he stated that 1:15,000 photographs allowed more detailed stratification. On this basis he recommended 1:20,000 as the smallest permissible scale for new aerial photography. This concurs with the statement by Harrison and Spurr (1955) that, for national forest inventories requiring accurate delineation of the principal commercial forest types, scales of 1:15,000 to 1:20,000 are usually recommended.
As regards smaller-scale photography, in the British colonies the scale has generally been around 1:30,000, occasionally larger, but since improved cameras became available the scale of 1:40,000 has been used (Miller 1957). However, photography taken primarily for forestry purposes for other agencies than for the Directorate of Overseas Surveys has usually been at scales of 1:15,000 to 1:20,000. For their investigations of forest cover types, E. C. Francis and Wood (1955) used the photographic scales 1:25,000 to 1:30,000, and Taylor and Stewart (1958) used 1:40,000. Through the agency of the Colombo Plan, complete photographic cover of Ceylon has recently been obtained on the scale of 1:40,000. The gross vegetation types can be recognized at this scale, although the leader of forest inventory work there, J.R.T. Andrews, preferred 1:20,000 as a minimum scale. Similarly, Rollet's stratification in Cambodia was done from 1:40,000 photographs. The forest inventory plan by Wheeler (1959) in Cambodia makes use of photographs mainly on the scale 1:40,000. However, 30 percent of land area has been photographed at the scale 1:10,000, and strips at 16 kilometer intervals of the rest of the country have also been photographed at this scale. D. A. Francis (1957) considered 1:20,000 photography the minimum for tree species identification, but found a smaller scale preferable at least in the initial stages of photo-interpretation work in the tropics. Subsequently Francis told the present author that he considered smaller-scale photography more efficient on the whole.
Large-scale (1:10,000 to 1:15,000) photography has been employed primarily in certain experimental inventories and recommended for more intensive studies. As a rule, however, tree species identification and volume assessment cannot be performed satisfactorily from photographs of even this large scale. As the principal use of the photographs is for classification of the area into cover types, and since this can usually be done from 1:30,000 to 1:40,000 photographs, this scale of photography is recommendable primarily for reasons of economy. To date, photography on this scale has been done and is available for large tropical areas which should be borne in mind in practical planning. Taken with new cameras and in the right conditions, such photography is a great help in the inventory of forests.
It does not mean, however, that large-scale photographs can be dispensed with in tropical forest surveys; these are required for special problems, and generally for intensive forestry purposes. They may be of considerable importance in the future.
The most important points deciding the selection of film for the inventory of tropical forests are the identification of tree species and forest types. Suitable film and filter combinations facilitate this work. The present situation in this respect can be summarized in brief.
The most important types of film for these purposes are panchromatic, infrared and color film, all of which meet somewhat differing quality requirements. Certain other film is mentioned in the literature - for example, spectrozonal film, claimed to have been developed by the Russians, may be worth considering (Hildebrandt 1957) - but so far, at least, these other types have not been used extensively. In certain phases, the possible uses of color film in the tropics have been viewed very optimistically (for example, D. A. Francis 1957), but no great importance can be attributed to color at present.
The material is expensive and the tonal differences on the film usually are fairly poor in the important green color zone.
Infrared film has been studied and employed a great deal. Its use in the identification of tree species is particularly appropriate when there are both broadleaved and coniferous trees in the forest. This is hardly the case in the tropics if at all, but infrared film has often been recommended anyway as it is supposed to have a better haze penetration than panchromatic film in given conditions.
The general opinion is, however, that a fast and good-quality panchromatic film with yellow filter is the most recommendable in the tropics. It gives the best reproduction of tonal differences. Price considerations also favor panchromatic film. More data could often be collected by the simultaneous use of two cameras, one loaded with infrared film, than by using panchromatic film alone. In practice, however, this method has seldom been applied because of the difficulties and the additional cost.
It was stated at the outset of this discussion of aerial photographs as a source of forest survey information in the tropics that for several reasons it would be desirable to do with as little ground work as possible. This was borne in mind in the subsequent analysis of the experience gained in the tropics of tree species identification, classifying forest cover types, and quantitative estimation of growing stock.
In some cases, tree species identification was remarkably successful, but on the whole the results were not adequate. Similarly, the quantitative estimation of growing stock directly from photographs seems to be somewhat inaccurate, though there is not very much experience of this point. On the other hand, several successful experiments show that important forest cover types can be discerned from aerial photographs, although checking and supplementation from the ground is often required. Problems other than the above have not been discussed point by point in the present paper, but conclusions can be drawn regarding them on the basis of general data on aerial survey.
It is interesting to note the extent to which the objectives of tropical forest inventory can be achieved primarily by air photo-interpretation.
Many kinds of data can be obtained on the surface area and its distribution. The forested area can usually be discerned, and even its distribution into productive and other forests. The former can be distributed by forest cover types. An indirect idea can be obtained, further, of site types, but the requirements on this point and on classes of silvicultural condition etc. - for example, regeneration - cannot be satisfied by photo-interpretation alone.
As to the growing stock, some idea of the total volume can perhaps be obtained by means of aerial photographs but the proportion of commercial species and diameter relations of the growing stock, for instance, are poorly illuminated. Trees in tropical high forests are highly liable to have defects, and these are quite invisible from aerial photographs (Dawkins 1958). Finally, no idea can be obtained of tree growth or removal from aerial photographs.
All told, the commonest main objective of the inventory of growing stock, determination of the allowable cut, is not achieved. Photo-interpretation alone seems adequate only for reconnaissance surveys of certain types, and they can sometimes be effected by aircraft or helicopter reconnaissance even without photography. But the rule must be emphasized that ground work cannot be avoided in tropical forest surveys.
The use of aerial photographs, however, has important advantages in these surveys. Above all, they enable stratification and arrangement of the ground work in the most efficient manner, and are an excellent tool in surface area assessment. Photographs are often most useful in practical survey work, showing roads, drainage, major topographical features and forest boundaries. Consequently, combined aerial and ground surveys seem to offer the best prospects for inventories.
BACKSTRÖM, HELMER & WELANDER, ERIK, 1948. En preliminär undersökning rörande möjligheterna att skilja olika trädslag på flygbilder [A preliminary investigation on the possibilities to distinguish different tree species on aerial photographs]. Svenska Skogsvfören. Tidskr., 48: 180-200.
BELOV, S. V. &; ABCYBASEV, E. S. 1957. [An investigation of the reflectivity of (the foliage of) tree species]. Bot. Z., 42 (4): 617-534. (In Russian) (Abstract in For. Abstr., 19:583).
BOON, D. A. 1955-56. Recent developments in photo-interpretation of tropical forests. Photogrammetria, Amsterdam, 12:382-386.
BROWNE, F. a. 1958. Study of tropical vegetation: Sarawak. In UNESCO. Study of tropical vegetation: proceedings of the Kandy symposium: 166-167.
CAHUSAC, A. B. 1957. Forest mapping from aerial photographs in Uganda. Entebbe, Uganda Forest Department. (Seventh British Commonwealth Forestry Conference).
COLWEEL, ROBERT. N. 1946. The estimation of ground conditions from aerial photographic interpretation of vegetation types. Photogram. Engng., 12: 161- 161
DAWKINS, H. C. 1958. The management of natural tropical high forest with special reference to Uganda. Oxford, Imperial Forestry Institute. Institute Paper No. 34.
DILLEWIJN, F. J. VAX. 1957. Sleutel voor de interpretatie van begroeingsvormen uit luchtsfoto's 1:40,000 van het Noordelijk deel van Suriname [Key for the interpretation of vegetation types on aerial photographs in northern Surinam, scale 1:40,000]. Dienst's Landebosbeheer Suriname, Paramaribo.
FARRER, R. P. Grown diameter in relation to girth and height for three tree species in Tanganyika. (Typescript in the Library of the Imperial Forestry Institute, Oxford).
FRANCIS, D.A. 1957. The use of aerial photographs in tropical forests. Unasylva 11, pp. 103-109.
FRANCIS, D.A. 1959. Forest inventory in the Sudan. Sudan Silva, pp. 49-56.
FRANCIS, D.A. 1960 Interim report to the Government of the Sudan on forest inventory. FAO Interim Report No. 59/10/7837, Rome.
FRANCIS, E. C. 1955. Interpretation of mangrove species from aerial photographs in North Borneo. (Typescript with sample photos in Directorate of Overseas Surveys, Tolworth, Surbiton, U.K.).
FRANCIS, E. C. & WOOD, G.H.S. 1954. Classification of vegetation in North Borneo from aerial photographs. Proc. 4th World For. Congr., Dehra Dun 1954, 3:623-629.
HAIG, IRVINE T. 1958. Tropical silviculture. Unasylva, 12:158-163.
HANNIBAL, L. W. 1952. Aerial forest photo-interpretation in Indonesia. FAO Asia-Pacific Forestry Commission. Document No. 75.
HARRISON, J.D.B. & SPURR, STEPHEN H. 1955. Planning a national forest inventory. Rome, FAO. (Mimeo).
HEINSDIJK DAMMIS. 1952. Forest photo-interpretation in Surinam. Photogram. Engng, 18:158-162.
HEINSDIJK, DAMMIS. 1955. Forest type mapping with the help of aerial photographs in the tropics. Trop. Woods (102): 27-46.
HEINSDIJK. DAMMIS. 1957-58. The upper story of tropical forests. Trop. Woods (107): 66-84; (108): 31-45.
HILDEBRANDT, G. 1957. Zur Frage des Bildmasstabes und der Filmwahl bei Luftbildautnahmen für forstliche Zwecke. Archiv für Forstwesen, 6:285-306.
HINDELY, E. & SMITH, J.H.G. 1967. Spectrophotometric analysis of foliage of some British Columbia conifers. Photogram. Engng, 23:894-895.
HOLLERWÖGER, F. 1954. Is there a correlation in the teak forests between crown diameter and the height of the trees with regard to the diameter on breast height? Journal of Scientific Research in Indonesia, 3:3-20.
HOWROYD, O. S. 1954. An investigation into commercial timber volume estimation from air photographs. (Typescript in the Library of the Imperial Forestry Institute, Oxford).
ILVESSALO YRJÖ 1960. On the correlation between the crown diameter and the stem of trees. Comm. Inst. Forest. Fenniae, 38 (2).
Jaarverslag van den Dienst Landsbosbeheer. 1957. Suriname.
LOETSCH, F. 1957a. Report to the Government of Thailand on inventory methods for tropical forests. Part One. Text. Part Two. Technical Appendices. Rome. FAO/ETAP Report No. 545.
LOETSCH, F. 1957b. A Forest inventory in Thailand. Unasylva, 11:174-180.
MERRITT, V. G. & RANATUNGA, M. S. 1959. Aerial photographic survey of the Sinharaja Forest. The Ceylon Forester, 4:103-156.
MILLER, R. G. 1957. The use of aerial photographs in forestry ire British Colonies. London, Colonial Office. Miscellaneous No. 528. (Seventh British Commonwealth Forestry Conference).
NAKAYAMA, HIROICHI. 1958. A study on forest survey from aerial photographs. Bull. of the Nagoya University Forests No. 1.
NYYSSÖNEN, AARNE. 1955. On the estimation of the growing stock from aerial photographs. Comm. Inst. Forest. Fenniae 46 (1).
PAELINCK,, P. 1958. Note sur l'estimation du volume des peuplements à limba (Terminalia superba) au Mayumbe, à l'aide des photos aériennes. Bull. agric. Congo beige, 49:1045-1054.
PAIJMANS, K. 1951. Een voorbeeld van interpretatie van luchtfoto's van oerwoud: het Malili-complex op Celebes [Interpretation of aerial photographs in virgin forest complex: Malili, Celebes]. Tectona, 111-136.
ROGERS, EARL J. 1958. Report of Working Group 4 (Foresters), Commission VII, International Society of Photogrammetry. Photogram. Engng., 24:603-616.
ROLLET, r, B. 1959. Rapport résumé d'activité pour la période du 15-2-1958 au 31-7-1959. Rome, FAO.
ROSAYRO, R. A. DE. 1958. Tropical ecological studies in Ceylon. In UNESCO. Study of tropical vegetation: proceedings of the Kandy symposium: 33-39.
SWELLENGREBEL, E.J.G. 1959. On the value of large scale aerial photographs in British Guiana forestry. Emp. For. Rev., 38:54-64.
TAYLOR, B. W. & STEWART. G. A. 1958. Vegetation mapping in the territories of Papua and New Guinea conducted by the CSIRO. In UNESCO. Study of tropical vegetation: proceedings of the Kandy symposium: 127-136.
WHEELER, P. R. 1959. Preliminary plan: forest survey of Cambodia. Cambodia, USOM.